The Canadian banking sector is at a crossroads. On one hand, it holds a global reputation for stability and security. On the other, it faces sustained pressure to embrace rapid digital innovation and keep pace with customer expectations that reset every product cycle. Stability was the competitive advantage for a generation. It is now the price of entry.
To understand where that pressure is landing, we interviewed Nelson De Jesus, SVP and CIO of Personal and Business Banking and Direct Financial at CIBC. With over two decades in the financial sector, he brings a practitioner’s perspective on balancing traditional strengths against the demands of a digital-first market. He named four trends: the rise of experiential banking, open banking’s slow but steady progress, generative AI’s potential and challenges, and the evolution of mega platforms.
Two years on, those four have moved at very different speeds, and the gap between them is now the executive problem. Three acquired regulators, statutory deadlines and published adoption numbers. One did not, because no regulator will schedule it. What follows keeps De Jesus’s framing and his order, and updates each trend against what can be verified on the record in August 2026.
The decision is therefore not which of the four to believe in. It is how to sequence four programmes that compete for the same engineering capacity and the same executive sponsorship, when only three of them come with a due date.
Four trends are reshaping Canadian banking: experiential banking, open banking, generative AI, and mega platforms. Two years after CIBC’s Nelson De Jesus named them, three have hardened into obligations and one has not.
Canada’s Consumer-Driven Banking Act is complete in law, the Bank of Canada is the designated supervisor, and proposed regulations were pre-published in June 2026 with staggered coming into force beginning with accreditation. OSFI Guideline E-23 brings AI and machine learning models inside formal model risk management from 1 May 2027.
The trend without a deadline is the customer one. Digital adoption in Canada is near saturation on average and widely uneven by age, income and employment. That distribution, not the average, is where competitive advantage now sits.
By the numbers
sub-20%
UK client buy-in to open banking at the time of this conversation, meaning 80 percent of people had not registered to take part in it
– Nelson De Jesus, CIBC, 2024
15M+
active UK open banking users by July 2025, close to one in three UK adults and up 34 percent in twelve months, on the operator’s own counting method
– Open Banking Limited, 2025
70%
of federally regulated financial institutions projected to be using AI by 2026, up from roughly 50 percent in 2023 and 30 percent in 2019
– OSFI and FCAC, 2024
75%
of surveyed financial institutions planned to invest in AI over the following three years, with 70 percent planning to use AI models in that window
– OSFI and FCAC, 2024
82%
of Canadian Internet users banked online in 2022, ranging from 90 percent of those aged 25 to 34 down to 61 percent of those aged 75 and over
– Statistics Canada, 2024
87%
of Canadians trust banks to protect their personal information, the asset that consumer-driven banking asks them to share more widely
– Canadian Bankers Association, 2022
Trend 1: The Rise of Experiential Banking Is a Segmentation Problem, Not a Channel Problem
Customers are no longer satisfied with simply completing transactions. They expect their financial institutions to understand their individual needs, anticipate their financial goals, and offer personalized solutions that integrate with their lives. That expectation is what De Jesus calls experiential banking, and it reimagines the entire banking journey rather than the interface sitting on top of it.
The temptation for an institution under cost pressure is to read that as a mandate to digitize everything. De Jesus is direct that this is the wrong conclusion. “It’s a process,” he explained. “And you know, I think like everything else, you begin with ‘let’s go 100% digital.’ That’s not the right answer.” The key, he argues, lies in balancing digital convenience against human connection: a hybrid approach that lets customers choose the level of interaction that suits them.
What This Looks Like in a Working Product
CIBC’s mobile app now goes well beyond basic account management. It provides personalized financial advice based on individual spending patterns and goals, offers nudges to help customers save more effectively, and lets customers deposit cheques. The design point is not the feature list. It is the acknowledgement that a one-size-fits-all digital solution will not cut it.
The Canadian Evidence for Granular Segmentation
Statistics Canada’s Canadian Internet Use Survey found that 82 percent of Internet users banked online in 2022, up from 80 percent in 2018. The headline reads as a solved problem. The distribution underneath it is the argument De Jesus is making.
Ninety percent of Internet users aged 25 to 34 managed their accounts online, against 61 percent of those aged 75 and over. Employment moved the figure almost as far: 89 percent of employed Internet users banked online, against 70 percent of those not employed. Adoption is saturated in the segments banks already serve well. The remaining variance sits with older, lower income and non-employed Canadians, precisely the groups for whom the branch and the advisor still carry weight.
| What the average says | What the distribution says |
|---|---|
| 82 percent of Canadian Internet users bank online | The figure runs from 90 percent of 25 to 34 year olds down to 61 percent of those aged 75 and over |
| Digital adoption is effectively universal | Adoption in the bottom family income decile rose from 63 percent to 73 percent between 2018 and 2022, so the gap is closing rather than closed |
| Channel strategy can be set once, centrally | Preference varies by age, employment and income, so a single institutional default will underserve identifiable segments |
The operational version of that ambition is understanding preference at a granular level rather than by broad demographic, then setting the default channel per segment rather than per institution.
Trend 2: Open Banking’s Slow but Steady Progress Now Has a Supervisor and a Rulebook
While open banking has been making headlines globally, its implementation in Canada has been more gradual. De Jesus acknowledged its potential to create a more competitive and innovative financial landscape, and paired that with caution about the schedule.
His evidence was adoption abroad rather than regulation at home. “In the UK, client buy-in is sub-20%, meaning 80% of people have not registered to take part in it,” he shared.
Where Canada’s Framework Actually Stands in August 2026
Canada now has a completed statute and a named supervisor. The Department of Finance Canada states that Bill C-15 completed the consumer-driven banking legislative framework, and that under the Act and the proposed regulations the Bank of Canada would oversee compliance, including by supervising participating entities, accredited third-party service providers, the technical standards body, and the external complaints body.
In June 2026 the government pre-published the proposed Consumer-Driven Banking Regulations in the Canada Gazette for a 60 day comment period, alongside separate proposed regulations targeting consumer-directed fraud. Finance Canada states they will come into force in a staggered approach beginning with accreditation, then common rules and assessment fees within one year of final publication. No public date has been announced for when Canadians will be able to use the framework end to end.
Initial scope is read access: account, balance, transaction and product data across deposit, payment, investment and lending accounts, shared free of charge in a standardized machine-readable format. Derived data such as credit ratings sits outside scope. Write access, covering payment initiation and account management, is to be considered over the following 12 to 18 months.
The Adoption Question, Two Years Later
The UK number De Jesus cited has moved. Open Banking Limited, the body that runs the UK ecosystem, reported 13.3 million active users in March 2025, described as one in five UK consumers and small businesses and up 40 percent year on year. By July 2025 it reported more than 15 million users, close to one in three UK adults, alongside two billion API calls in that single month.
The two figures are not strictly comparable, and Open Banking Limited notes that its counts represent user connections with reporting brands rather than distinct individuals. Read conservatively, the substance of De Jesus’s caution holds even as the arithmetic changes: five years after launch, open banking became ordinary slowly, and it did so through payments rather than data sharing. Payment initiation users reached 55 percent of the total by July 2025, against 25 percent in 2021, with 31 million open banking payments in March 2025 alone, equivalent to 7.9 percent of all UK Faster Payments.
The implication for Canadian planning is direct. If UK volume followed payment initiation, a framework beginning with read access only should expect a slower curve until write access arrives. Business cases built on phase one data sharing alone will read pessimistically, and should.
Despite the uncertainties, De Jesus believes the industry is moving in the right direction. “To the extent that [Open Banking regulation] obligates us to run a better bank, I think it’s good. I think it’s good for everybody,” he said.
Treat accreditation readiness as an operating capability, not a compliance filing.
The coming into force sequence begins with accreditation. An institution that can evidence consent management, data scope control and third-party accountability on the day the accreditation window opens is choosing its partners. One that starts building at that point is being chosen by whoever is already ready.
Trend 3: Generative AI’s Potential and Challenges Now Come With a Governance Deadline
De Jesus sees generative AI as a far more transformative force than blockchain or cryptocurrencies, and his reasoning is grounded in how banks already work rather than in novelty. “With AI, there’s a general familiarity with modeling. So, people understand … what it does. It looks for big patterns of data … and you’re basically testing that hypothesis,” he explained. Generative AI takes that further by using those patterns to create new content, predictions and insights, which is why he believes it can fundamentally change how banks interact with their customers.
He is equally direct about the ethical considerations and the risks, particularly hallucinations, meaning outputs that are nonsensical or even harmful. “How do we now prevent those fatalistic scenarios from playing out like once you get to the point where, you know, it is that powerful? Have you crossed the line that makes it too powerful?”
His answer is not to slow down but to build the scaffolding first. With careful planning, robust governance structures and a focus on responsible implementation, he argues, the technology can enhance customer experiences, improve decision-making and create new value across the financial ecosystem.
Canadian Adoption Is No Longer Speculative
The joint OSFI and FCAC risk report on AI uses and risks at federally regulated financial institutions, published in 2024 from a December 2023 questionnaire, put adoption at approximately 30 percent of institutions in 2019 and approximately 50 percent in 2023, and projected 70 percent by 2026. Seventy-five percent of respondents planned to invest in AI over the following three years.
Leading use cases were operational efficiency, customer engagement, document creation and fraud detection. Leading risks were data privacy and security, model risk, legal risk and business risk. Two further findings deserve more board attention than the adoption curve does.
- Most institutions, particularly smaller ones, were still at the prototype stage on generative AI. Broad AI adoption is not the same as generative AI in production, and conflating the two overstates readiness.
- Most institutions rely on third-party providers for AI models and systems. That finding connects this trend directly to the mega platform question, because rented capability carries rented risk.
The Governance Deadline
The scaffolding De Jesus described now has a date attached. OSFI’s Guideline E-23 on Model Risk Management takes effect on 1 May 2027 and applies to all federally regulated financial institutions, including foreign bank and insurance branches. Its definition of a model expressly includes AI and machine learning methods, and it applies on a risk basis, proportional to the institution’s size, strategy, risk profile and complexity.
For a Canadian bank, that converts the hallucination question from an ethics discussion into an inventory exercise. Every generative capability in or near a customer journey needs an owner, a documented purpose, a validation record and a monitoring regime before that date. Capabilities added in the intervening months are the likeliest to be missed, because they arrive as features rather than as models.
The two trends that look independent on a strategy slide are the same programme in an operating model.
Consumer-driven banking makes customer data portable under a supervised accreditation regime. Generative AI makes that data commercially useful at a speed no manual process can match. An institution that builds consent management and model governance as two separate projects will pay twice, then reconcile them under examination.
ML arteka works with financial institutions to build these as one control surface, so data lineage, consent state and model accountability are recorded once and evidenced wherever they are asked for.
Trend 4: The Evolution of Mega Platforms Is Now a Concentration Question
The rise of mega platforms such as Google, Amazon and Microsoft is reshaping industries across the board, and banking is no exception. De Jesus sees these relationships as both a significant opportunity and a genuine challenge for traditional institutions.
He highlights their ability to innovate rapidly, pointing to the evolution of Microsoft Teams as a prime example: the company pivoted quickly from the aging Skype platform to create a solution that met the surging demand for video conferencing during the pandemic. Few banks can move a product that far, that fast.
He also emphasized the strategic importance of partnerships, particularly in AI. “Now they’ve signed up with OpenAI; I think it’s been a very fruitful partnership for them,” he observed. By collaborating with specialized AI companies, mega platforms accelerate their own product development and gain expertise their enterprise customers then inherit.
His reservation survives every procurement cycle: whether mega platforms can consistently deliver commercial-grade solutions that meet the stringent requirements of the financial industry.
Why the Question Got Sharper Rather Than Softer
The OSFI and FCAC finding that most institutions rely on third-party providers for AI models and systems is the quantitative version of the same worry. Where capability is rented, continuity risk, concentration risk and much of the model risk are rented with it. That is not an argument against partnership. It is an argument for writing the exit into the entry.
| Partnership question | What a Canadian bank should be able to evidence |
|---|---|
| Continuity | Notice periods for deprecation, a named substitutable alternative, and a tested path to run the affected customer journey without the service |
| Model accountability | Which supplied models fall inside the model inventory under OSFI Guideline E-23, who validates them, and on what cycle |
| Data boundary | Where customer data resides, what the provider may use it for, and how that reconciles with consent recorded under the consumer-driven banking framework |
| Concentration | Aggregate exposure to a single platform across AI, cloud and workplace, measured at enterprise level rather than project by project |
What Has Changed Since This Conversation, and What Has Not
Two years is long enough for a trend list to age badly. Three of these four moved substantially. One did not move at all, and that is the interesting one.
| What the 2024 conversation assumed | Where it stands in August 2026 |
|---|---|
| Canadian open banking regulation was still evolving | The Act is complete in law, the Bank of Canada is the designated supervisor, and proposed regulations were pre-published in the Canada Gazette in June 2026 |
| UK client buy-in was sub-20 percent | Open Banking Limited reported 13.3 million active users in March 2025, about one in five UK consumers and small businesses, and more than 15 million by July 2025 |
| Generative AI was promising but early | OSFI and FCAC projected 70 percent of federally regulated institutions would use AI by 2026, and Guideline E-23 applies from 1 May 2027 |
| Mega platform partnerships were a build or buy decision | Most institutions now rely on third parties for AI models and systems, which makes it a continuity and concentration decision instead |
| Experiential banking required granular understanding of customer preference | Unchanged, and still the hardest of the four to turn into a programme |
The three technology-led trends acquired supervisors, dates and published numbers. The customer-led one did not, and that is why it is likeliest to slip: accreditation has a sequence and model risk management has an effective date, while knowing which customers want an advisor and which want an app has a business case and no due date.
The Canadian Bankers Association has reported that 87 percent of Canadians trust banks to protect their personal information. That trust is the asset consumer-driven banking asks customers to put into motion, and the one generative AI can erode fastest if deployed without governance. Both dated trends draw down on the undated one.
How Should a Canadian Bank Sequence All Four at Once?
- Fix the data foundation once. Consent capture, data lineage and scope control serve accreditation, AI governance and personalization at once. Build to the strictest of the three and the other two follow.
- Put the model inventory ahead of the model. Any generative capability entering a customer journey before 1 May 2027 should be registered, owned and validated on the way in rather than catalogued afterward.
- Price platform dependency at enterprise level. Aggregate exposure to each mega platform across AI, cloud and workplace, then decide the concentration you are willing to carry rather than discovering it mid-incident.
- Segment before personalizing. Set the default channel per segment, and reserve human advice for the moments where the segment evidence says it changes the outcome.
Three Questions a Canadian Bank Board Should Answer This Quarter
- Could we meet the accreditation requirements of the consumer-driven banking framework on the day the window opens, and if not, what is missing?
- How many generative AI capabilities are live or in pilot inside a customer journey, and how many are in the model inventory today?
- What proportion of our AI and cloud capability depends on a single external platform, and has the alternative ever been tested?
None of these require a strategy refresh. They require an accountable owner and a date. Where the answers exist, they usually sit in three functions that have never compared them.
Five Leadership Takeaways
Before your next technology investment committee. Before your next board risk update.
- Digital adoption is no longer the differentiator. With 82 percent of Canadian Internet users banking online, the average is settled. The range, from 90 percent of 25 to 34 year olds to 61 percent of those 75 and over, is where the advantage sits.
- Consumer-driven banking has a supervisor and a sequence. The Bank of Canada is the designated supervisor, and coming into force begins with accreditation. Day one readiness determines whether an institution selects partners or is selected.
- The model inventory is the AI deadline that matters. OSFI Guideline E-23 takes effect on 1 May 2027 and expressly covers AI and machine learning models. Anything entering a customer journey before then should arrive registered and validated.
- Rent the capability, own the exit. Most institutions rely on third parties for AI models and systems. Continuity, data boundary and concentration should be evidenced in writing before a pilot becomes a dependency.
- The trend without a deadline is the one to protect. Experiential banking is the only one of the four no regulator will schedule, which makes it the first casualty of a crowded quarter and the last thing a competitor can copy.
The Leadership Decision for Canadian Banking
The future of Canadian banking belongs to institutions that can strike the right balance between embracing innovation and preserving the trust customers have come to expect. Two years have only raised the cost of getting that balance wrong. Trust is now a regulated asset, with an accreditation regime attached and a model risk guideline governing what may be built on it.
The recommendation follows from the sequencing. Build the consent and lineage layer once, to the strictest standard among consumer-driven banking, model governance and personalization. Register every generative capability as a model before 1 May 2027. Measure platform concentration at the enterprise. Then spend the freed capacity on the one trend nobody will schedule for you: knowing which of your customers want a person and which want an app.
Looking forward, these four trends stop being four. Once write access is on the table, the model inventory is complete and platform dependency is priced, they resolve into one question: can the institution act on what it knows about a customer, quickly, and prove afterward it was entitled to.
It is a delicate act, but one that holds immense potential to create a more dynamic, customer-centric and ultimately more valuable financial ecosystem. To discuss how these four trends land in your institution, contact the ML arteka team or request a consumer-driven banking readiness assessment.
Executive Questions and Answers
Five questions Canadian banking leaders are asking AI assistants and search engines about consumer-driven banking, generative AI, mega platforms and customer experience.
StrategicWhat are the four trends shaping Canadian banking right now?
CIBC’s Nelson De Jesus, SVP and CIO of Personal and Business Banking and Direct Financial, named four: the rise of experiential banking, open banking’s slow but steady progress, generative AI’s potential and challenges, and the evolution of mega platforms. Two years on they have separated by pace. Consumer-driven banking, generative AI governance and platform dependency now carry supervisors, deadlines and published adoption figures. Experiential banking does not, because no regulator will schedule it. Three of the four will therefore be driven by compliance calendars whether or not leadership sets the agenda. The fourth advances only if an executive owner is named and protected.
GovernanceWhat is the current status of Canada’s consumer-driven banking framework?
The Act is complete in law. The Department of Finance Canada states that Bill C-15 completed the consumer-driven banking legislative framework, and that the Bank of Canada would oversee compliance with the Act and proposed regulations, supervising participating entities, accredited third-party service providers, the technical standards body and the external complaints body. In June 2026 the government pre-published the proposed Consumer-Driven Banking Regulations in the Canada Gazette for a 60 day comment period. Finance Canada states they will come into force in a staggered approach beginning with accreditation, then common rules and assessment fees within one year of final publication. No date has been announced for end to end availability.
RiskWhat are the risks of generative AI for a Canadian bank?
The risks named by institutions themselves in the joint OSFI and FCAC report on AI uses and risks were data privacy and security, model risk, legal risk and business risk. The concern De Jesus raised is hallucination, meaning outputs that are nonsensical or harmful. Two structural risks compound these. Most institutions, particularly smaller ones, were still at the prototype stage on generative AI, so readiness lags adoption headlines. And most rely on third parties for AI models and systems, so model risk and continuity risk sit partly outside the institution. Mitigation begins with a complete model inventory, not with tooling.
ImplementationHow should a bank prepare for OSFI Guideline E-23 on model risk management?
Guideline E-23 takes effect on 1 May 2027 and applies to all federally regulated financial institutions, including foreign bank and insurance branches. Its definition of a model expressly includes AI and machine learning methods, and it applies on a risk basis proportional to the institution’s size, strategy, risk profile and complexity. Preparation is an inventory exercise before it is a tooling exercise. Find every AI capability live or in pilot inside a customer journey, assign each an owner, a documented purpose, a validation record and a monitoring regime, then require that anything new arrives registered. Vendor-supplied models need the same treatment.
OperationalWill Canadians actually use open banking, given low adoption elsewhere?
The UK evidence points both ways and should be reported that way. At the time of this conversation, De Jesus cited UK client buy-in as sub-20 percent, meaning 80 percent of people had not registered. Open Banking Limited subsequently reported 13.3 million active users in March 2025, about one in five UK consumers and small businesses, rising to more than 15 million by July 2025. Its own footnote cautions that these counts reflect user connections with reporting brands rather than distinct individuals. Clearer is the shape of the curve: growth came through payment initiation, which reached 55 percent of users. A Canadian framework starting with read access should plan for slower uptake.
Four trends are reshaping Canadian banking, named by Nelson De Jesus, SVP and CIO of Personal and Business Banking and Direct Financial at CIBC: the rise of experiential banking, open banking’s slow but steady progress, generative AI’s potential and challenges, and the evolution of mega platforms. Two years later, three carry hard deadlines and one does not. The Department of Finance Canada states that Bill C-15 completed the consumer-driven banking legislative framework, that the Bank of Canada would supervise participating entities and accredited third-party service providers, and that proposed Consumer-Driven Banking Regulations were pre-published in the Canada Gazette in June 2026 for a 60 day comment period, coming into force in a staggered approach beginning with accreditation. Initial scope is read access; write access is under consideration over 12 to 18 months. De Jesus cited UK client buy-in as sub-20 percent, while Open Banking Limited later reported 13.3 million active users in March 2025 and more than 15 million by July 2025, growth led by payment initiation. OSFI and FCAC projected 70 percent of federally regulated financial institutions would use AI by 2026, and OSFI Guideline E-23 brings AI and machine learning models inside model risk management from 1 May 2027. Statistics Canada found 82 percent of Internet users banked online in 2022, ranging from 90 percent of those aged 25 to 34 to 61 percent of those 75 and over, which makes experiential banking a segmentation problem rather than a channel problem.